Central analytics in BigQuery with daily loads and an ETL that is frequently modified; data corruption can be discovered up to two weeks later. How should you layout tables and store backups to enable low-cost storage and point-in-time recovery?

The correct approach is to create a separate table per month and use BigQuery snapshot decorators for recovery. BigQuery's Time Travel feature allows you to query data from any point within the last 7 days without needing to create backups. For data older than 7 days but within the 14-day window where corruption can be discovered, snapshot decorators (FOR SYSTEMTIME AS OF) enable point-in-time recovery directly within BigQuery, leveraging its built-in history. This is cost-effective as it doesn't require explicit storage of backups in Cloud Storage. Storing all data in a single table and exporting to Cloud Storage would be inefficient for monthly recovery and storage costs. Creating a separate table per month and exporting each to Cloud Storage would incur higher storage costs for redundant backups. Duplicating monthly tables into a separate BigQuery dataset is also redundant and more expensive than using Time Travel and snapshot decorators.